With AI adoption accelerating across enterprises, what do you view as the top information security challenge that leaders should address? How can CISOs, VPs and Director’s align governance, risk and security investments to enable AI innovation without creating new exposures?

One of the biggest security challenges with AI adoption is ensuring the integrity, confidentiality, and responsible use of the data and Machine Learning models behind it. As organizations move quickly to innovate, risks such as data poisoning, ML model manipulation, and ungoverned “shadow AI” across the organization can undermine trust if not addressed soon. Security leaders should work to embed AI governance into existing frameworks, drawing on standards like the NIST AI RMF. Prioritizing investments in data protection, data privacy, model monitoring, and AI red teaming can strengthen resilience without slowing down innovation. By approaching AI as both a valuable asset and a potential risk vector, leaders can put the right guardrails in place to support adoption that is safe, scalable, and sustainable.

Before adopting AI, it is important to be clear about what it can truly solve and what it cannot. This helps set realistic expectations, measure ROI properly, and avoid unnecessary risk or disappointment when the technology is not addressing a real business problem.

AI also brings a lot of uncertainty. Its performance can be hard to predict, regulations are still evolving, and outcomes can be biased or unexpected. To manage this, organizations need a cross functional team that brings together : technical, business, legal, and leadership expertise. 

They also need a governance model with clear roles and decision processes, plus the ability to anticipate regulatory changes before they become constraints. Just as important the data must be clean, reliable, and compliant. 

Data verification from AI results is a top priority but requires expertise in the answers developed so I would say a referendum of fact verification. There's possibility that results from AI is incorrect or simply wrong due to different reasons especially since it is new to the public on a large scale. Only time will tell how reliable AI sourced adoption will incur in the next five to ten years. 

Balancing 1) how pervasive and inevitable it is, 2) the general low awareness and understanding of it's risks, limitations, biases and applicable use cases for different types of AI within most organisations, and 3) the need for good governance, policies, education and tooling that enables and protects the organisation's data and information, whilst recognising the time it takes to implement each of these controls.

It's very much a case of if you're not already moving at pace to stay ahead or at least keep up, you're already behind. In some ways, the most 'risk averse' and 'risk accepting' organisations are both equally vulnerable with those that can find the sweet spot in between by balancing these things most likely to navigate it successfully.

Implement a DLP solution to ensure privacy and security of data are preserved.